Georgia Institute of Technology
Computational Imaging-Based Prognostic and Predictive Biomarkers for Oropharyngeal Cancer
Abstract
dc:description.abstractIt is estimated that 54,540 adults in the United States will be diagnosed with oropharyngeal squamous cell carcinoma (OPSCC) in 2024, 70% of which will be related to human papillomavirus (HPV+). HPV+ OPSCC differs from HPV-negative (HPV-) OPSCC, which is associated with tobacco and alcohol use, in terms of disease aggressiveness and response to treatment. These differences have led to different treatment paradigms, which requires separate risk stratification strategies for the two cancer entities to guide treatment planning. Chemoradiation is a common therapeutic regimen for p16+ OPSCC. However, adding chemotherapy can increase treatment-related toxicity, a particular concern for patients who are at low risk of disease recurrence. This underscores the need to develop predictive biomarkers to identify low-risk OPSCC patients who may not benefit from chemotherapy and could avoid such toxic treatments without compromising their outcomes. The prognosis of p16+ OPSCC is highly influenced by the characteristics of both the primary tumor (PT) and metastatic cervical lymph nodes (LN). Although AI-based prognostic biomarkers focusing on pathology and radiology have been developed independently for head and neck cancers, the integration of PT and metastatic cervical LN across radiology and pathology remains unexplored. The objective of this thesis is to develop and validate four computational workflows. Firstly, we develop a prognostic tool for OPSCC risk stratifications taking HPV status into account. Secondly, we developed an predictive CT biomarker to identify early-stage OPSCC candidates who will not benefit from chemotherapy, thus suitable for treatment tailoring. Thirdly we investigated a deep learning model integrating PT and metastatic cervical LN to risk-stratify OPSCC patients into low- vs high-risk groups based on disease-free survival, overall survival and locoregional failure. Finally, we investigated the prognostic value of an multimodal AI approach, termed the swin-transformer-based multimodal and multi-region data fusion framework (SMuRF), to predict outcomes of OPSCC.
Degree
thesis:*- Level thesis:degree_level
- Doctoral
- Department dc:contributor.department
- Biomedical Engineering (Joint GT/Emory Department)
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Song, Bolin
- Advisor dc:contributor.advisor
-
- Madabhushi, Anant
- Committee members dc:contributor.committeemember
-
- Saba, Nabil F.
- Coskun, Ahmet F.
- Wang, May D.
- Sinha, Saurabh
Subjects
dc:subject × 5Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/76920